H-Scale: Hessian-Guided Scale Refinement for NVFP4 Sub-Byte LLM Inference

• URL: http://arxiv.org/abs/2608.28113v1
• Date: Fri, 28 Aug 2026 09:22:17 GMT
• Title: H-Scale: Hessian-Guided Scale Refinement for NVFP4 Sub-Byte LLM Inference
• Abstract summary: H-Scale is a lightweight post-processing method for NVFP4 per-group scale refinement.<n>It is designed as a drop-in replacement for RTN-style scale selection in diverse NVFP4 pipelines.<n> experiments show that H-Scale generally improves a broad range of NVFP4 baselines.
• Score: 27.189669587707858
• License: http://creativecommons.org/publicdomain/zero/1.0/
• Abstract: The NVIDIA Blackwell architecture, with native support for the ultra-fine-grained NVFP4 format, opens new opportunities for accelerating large language model (LLM) inference. NVFP4's micro-block design, such as a group size of 16, offers strong representational flexibility for capturing local weight distributions and isolating outliers, but it also introduces a large and highly sensitive space of per-group scaling factors. Existing post-training quantization (PTQ) methods primarily focus on refining quantized weight values, leaving this scale-selection step underexplored. To address this gap, we propose \textbf{H-Scale}, a lightweight post-processing method for NVFP4 per-group scale refinement. Instead of minimizing plain weight reconstruction error, H-Scale selects hardware-valid group scales using a diagonal second-order proxy derived from calibration activations, thereby targeting layer output perturbation more directly. It is designed as a drop-in replacement for RTN-style scale selection in diverse NVFP4 pipelines, requires only modest offline calibration, and introduces strictly zero overhead at inference time. Under a fixed evaluation protocol, experiments on mainstream LLMs show that H-Scale generally improves a broad range of NVFP4 baselines and brings several variants closer to the BF16 reference.

Related papers

• FOCUS: FP4 Optimization via Coupled-Relaxation and Dual-Granularity Scaling [23.78963716646984]
  FOCUS is a post-training quantization framework with end-to-end scale learning for FP4 optimization.<n>It relaxes the tight coupling between quantization and dequantization scales with a learnable full-precision coefficient.<n>Experiments show that FOCUS achieves state-of-the-art FP4 accuracy under both MXFP4 and NVFP4 formats.
  arXiv  Detail & Related papers  (2026-08-03T07:54:56Z)
• ScaleSweep: Accurate NVFP4 Post-Training Quantization of LLMs via Block Scale Initialization [40.425536766574844]
  We propose ScaleSweep as a simple and efficient scale optimization method for NVFP4 quantization.<n> Experiments on Llama and Qwen models demonstrate that ScaleSweep consistently improves quantization performance.<n>In particular, under aggressive end-to-end quantization of weights, activations, KV cache, and query states, ScaleSweep preserves more than 93% of the full-precision performance.
  arXiv  Detail & Related papers  (2026-05-30T08:26:16Z)
• SOAR: Scale Optimization for Accurate Reconstruction in NVFP4 Quantization [19.022444007775896]
  NVFP4 has recently emerged as an efficient 4-bit microscaling format for large language models.<n>Existing methods often yield suboptimal performance due to inflexible scale selection and the coupled treatment of quantization and dequantization scales.<n>We propose Scale Optimization for Accurate Reconstruction (SOAR), a novel post-training quantization framework that improves the accuracy of NVFP4 quantization.
  arXiv  Detail & Related papers  (2026-05-12T15:13:18Z)
• FAAR: Format-Aware Adaptive Rounding for NVFP4 [10.642666144689787]
  Large language models (LLMs) on edge devices require extremely low-bit quantization.<n>Existing quantization methods fail to account for the non-uniformity of the NVFP4 numerical grid.<n>We propose Format-Aware Adaptive Rounding (FAAR), a learnable rounding strategy tailored for the NVFP4 format.
  arXiv  Detail & Related papers  (2026-03-23T06:29:01Z)
• Bridging the Gap Between Promise and Performance for Microscaling FP4 Quantization [77.67818998672516]
  We present the first comprehensive study of MXFP4 and NVFP4 for post-training quantization.<n>We introduce Micro-Rotated-GPTQ (MR-GPTQ), a variant of the classic GPTQ quantization algorithm.<n>We show that MR-GPTQ matches or outperforms state-of-the-art accuracy.
  arXiv  Detail & Related papers  (2025-09-27T09:22:21Z)
• Optimizing Large Language Model Training Using FP4 Quantization [73.55459961002371]
  Quantized training presents a promising solution by enabling low-bit arithmetic operations to reduce costs.<n>This work introduces the first FP4 training framework for large language models (LLMs)
  arXiv  Detail & Related papers  (2025-01-28T18:04:50Z)
• GAQAT: gradient-adaptive quantization-aware training for domain generalization [54.31450550793485]
  We propose a novel Gradient-Adaptive Quantization-Aware Training (GAQAT) framework for DG.<n>Our approach begins by identifying the scale-gradient conflict problem in low-precision quantization.<n>Extensive experiments validate the effectiveness of the proposed GAQAT framework.
  arXiv  Detail & Related papers  (2024-12-07T06:07:21Z)
• Pushing the Limits of Large Language Model Quantization via the Linearity Theorem [71.3332971315821]
  We present a "line theoremarity" establishing a direct relationship between the layer-wise $ell$ reconstruction error and the model perplexity increase due to quantization. This insight enables two novel applications: (1) a simple data-free LLM quantization method using Hadamard rotations and MSE-optimal grids, dubbed HIGGS, and (2) an optimal solution to the problem of finding non-uniform per-layer quantization levels.
  arXiv  Detail & Related papers  (2024-11-26T15:35:44Z)
• FlatQuant: Flatness Matters for LLM Quantization [58.28221892035609]
  We propose FlatQuant, a new post-training quantization approach that enhances the flatness of weights and activations.<n>Our approach identifies optimal affine transformations for each linear layer, calibrated in hours via a lightweight objective.<n>It achieves less than 1% accuracy drop for W4A4 quantization on the LLaMA-3-70B model, surpassing SpinQuant by 7.5%.
  arXiv  Detail & Related papers  (2024-10-12T08:10:28Z)
• SliM-LLM: Salience-Driven Mixed-Precision Quantization for Large Language Models [63.118592279833656]
  Post-training quantization (PTQ) is an effective technique for compressing large language models (LLMs)<n>We propose SliM-LLM, a salience-driven mixed-precision quantization framework that allocates bit-widths at the group-wise.<n> Experiments show that SliM-LLM achieves superior performance across various LLMs at low bit-widths.
  arXiv  Detail & Related papers  (2024-05-23T16:21:48Z)
• LLM-FP4: 4-Bit Floating-Point Quantized Transformers [38.23587031169402]
  We propose LLM-FP4 for quantizing both weights and activations in large language models (LLMs) down to 4-bit floating-point values. Compared to integer quantization, floating-point (FP) quantization is more flexible and can better handle long-tail or bell-shaped distributions. Our method, for the first time, can quantize both weights and activations in the LLaMA-13B to only 4-bit and achieves an average score of 63.1.
  arXiv  Detail & Related papers  (2023-10-25T17:59:32Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.